Let's start with a demo

Motion detection with Background Subtraction

• Store a model of the "background"
• Compare new images to this background model
• Infer that large differences are caused by motion

To illustrate an example of this, consider the following background model image:

Next, we take our latest frame,

Then blur it and convert it to black and white so it can be compared to our background image.

Taking the difference between the two frames yields the following "delta" image:

Next, we threshold (or binarize) the image, by converting all pixels above/below a certain threshold to white or black.

Dilation is then performed to close any gaps. Dilation involves moving a sliding window (kernel) over an image, and updating a given pixel's value based on the surrounding pixels.

if white_blobs == big:
alert()

You can see how all this is layered together in the demo below.

• slack is awesome
• slack has some amazing APIs
• slack works on mobile and desktop
• I don't need to code a GUI

Interaction through slash commands

entering /pycam_on in slack sends a post request to a Flask function

@app.route('/pycam_on', methods=["POST"])
@slack_verification()
def pycam_on():
    """Turn on the pycam process.

    Returns:
        str: Response to slack
    """
    if utils.redis_get('camera_status'):
        response = 'Pycam is already running'
    else:
        utils.redis_set('camera_status', True)
        response = "Pycam has been turned on"
    return response

More slash commands

What if I don't want to manually turn the system on/off?

Turns out my router has an API..

import requests

# Create a new requests session
session = requests.session()

# Call the login API with my credentials
response = session.post('<router_login_api_url>', LOGIN_DATA)

# Update my cookie with the new, authenticated user_id
user_str = response.headers['Set-Cookie']
user_id = user_str.split(';')[0].split('userid=')[1]
new_cookie = BASE_COOKIE.format(USERNAME, PWD, user_id)
headers['Cookie'] = new_cookie

# Call api and get connected devices
response = session.get('<get_connected_devices_api_url>', headers=headers)
devices = response.json()
connected_devices = [d['hostName'] for d in devices]

But how do you make sure all that stuff is running?

Enter glances.

do_deep_learning()
download pre-trained image classifier model

import cv2

# Load model
net = cv2.dnn.readNetFromCaffe(proto_file, caffe_model_file)

# Get classifications
net.setInput(img)
classifications = net.forward()

Results

If you actually care about home security...

you should probably buy an actual home security system

More important things than home security

• Are there leftovers in the company kitchen?
• Did my neighbors dog shit on my lawn?
• Did my roommate take my beer?

Thank you python community...

you allowed me to build a home security system with:

• no computer science background
• no idea how to build APIs or a web app
• no idea how motion detection works